Jump to content

Wikipedia talk:Correct typos in one click

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia
(Redirected from Wikipedia talk:CTIOC)

Toolforge tool

[edit]

Hello dear contributors, following some feedbacks that accummulated over the years (and the fact that I currently don't have a job), I have created a new toolforge tool:

https://typos.toolforge.org/

Few advantages that I had in mind during the creation of the tool:

  • verify each typo still exists in article before showing it to editor.
  • only show one typo at a time
  • added the skip button, in addition to dismiss button, so now you can easily skip typos that you don't want to deal with and don't want to dismiss as non typos. The skipped words will eventually come back up once I change their db status from skipped to new, but I will have some data about them and who skipped what. Maybe in the future I will implement eternal skip so that each editor won't have to see words he skipped ever again.
  • prevent case where the context is no longer the same in the article, hence failing the replacement. Now the context is taken from the article when the typo is loaded by the editor.
  • every typo that is served to an editor is marked as served, so it won't appear to another editor. After we finish the words and have some pause, I will mark the served words as new.
  • bigger buttons on mobile view and no need to expand paragraphs.
  • less "noise" to the project namespace, the only edits will be on the main namespace, and the counting will be done through the tags system (Special:Tags), it also allows following the recent edits in the project ([1]).

Please let me know what you think, every feedback is crucial for the success of the new tool and for the expansion of the project to new sister projects and languages.

It's an opportunity for me to thank the top contributors of the project (Quarry):

Uziel302 (talk) 06:26, 15 October 2022 (UTC)[reply]

I am currently fixing a bug User:Qwerfjkl reported, where his edit was registered under my name. Will update once it is solved. Uziel302 (talk) 07:03, 15 October 2022 (UTC)[reply]
Thank you! I won't have time to try it out today, but that sounds like a great series of improvements. However, I'm currently getting error 503 (No webservice) on https://typos.toolforge.org/. Certes (talk) 08:48, 15 October 2022 (UTC)[reply]
@Certes, I presume it's due to the fix mentioned above. It was working when I tried it.
@Uziel302, what will happen if the edit returns an error message e.g. editconflict, ratelimit, page protected etc.? — Qwerfjkltalk 09:13, 15 October 2022 (UTC)[reply]
Certes, User:Qwerfjkl, it seems to work fine now, tested with multiple users. As of errors, for now they will show up in the status line, as "[object Object]" after "processing" status. At least this is what I get with my bot account that is blocked on English Wikipedia. I will change it to show more relevant message. Uziel302 (talk) 16:10, 15 October 2022 (UTC)[reply]
@Uziel302, can you increase the amount of context shown (significantly)? There's a lot more space for it now. — Qwerfjkltalk 16:19, 15 October 2022 (UTC)[reply]
@Qwerfjkl, I increased context and changed the regex a bit (github) Uziel302 (talk) 17:25, 15 October 2022 (UTC)[reply]
@Uziel302, strange bug here where it treated part of a word as a word. Also, I'm not sure how the tool copes with multiple (identical) typos in the same article.
Should something be done about the current project page & subpages, or is it too soon? — Qwerfjkltalk 18:16, 15 October 2022 (UTC)[reply]
@Qwerfjkl, I added requirement for space before the word, like I require in the scan. I verified that the right edit is made - [2]. If the typo appears multiple times, it will take the first match. The next scan will catch it again so that it also will get fixed.
As of current project, I prefer not to touch it as long as we still dealing with bugs. Uziel302 (talk) 18:57, 15 October 2022 (UTC)[reply]
@Uziel302, Another bug: for this edit, it didn't show any context. — Qwerfjkltalk 19:45, 15 October 2022 (UTC)[reply]
This commit should solve it. Uziel302 (talk) 14:44, 16 October 2022 (UTC)[reply]
@Uziel302, a minor request, but could you use → instead of -> in the summary and => on the tool page?

Also, the text currently appears like:
do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut typo
typo ad minim veniam, quis nostrud exercitation ullamco

It would be nice if you could somehow squeeze in the suggested replacement somewhere, so I don't have to glance back up to check the correction.

Thanks for your hard work on this project, and quick response to the big reports! — Qwerfjkltalk 19:48, 16 October 2022 (UTC)[reply]
The suggested replacement now appears in the context. Uziel302 (talk) 12:42, 18 October 2022 (UTC)[reply]

@Uziel302, I have tried the new tool, which mostly seems to work well. A couple of errors had no content, breds in Swiss Warmblood and in Cleveland Bay. Also there is no type button. The tool probably needs some explanatory text, especially a note that corrections are suggested by a script and are not always accurate. I miss some aspects of the existing pages, where I can choose obvious corrections first, also words from foreign languages I am familiar with such as Dutch or Swedish. I can then return to the corrections that need a bit more work. TSventon (talk) 10:24, 16 October 2022 (UTC)[reply]

@TSventon, regarding "type", you can just change the word - it's in an input box. — Qwerfjkltalk 12:48, 16 October 2022 (UTC)[reply]
@TSventon, added note about accuracy. The missing context bug was fixed.
What is exactly "choose the obvious corrections first"? In my head it means that you scroll many corrections that require time. For this reason I made the skip button, it means "I don't want to deal with this typo now, but don't dismiss it as non typo". The skipped words will come up again eventually. In a way, I understand the advantage of a long list that you can shortly skim, but I don't want to recreate this experience in the new tool, it's messy and doesn't support real time verification that typo still exists. We can keep using the old project alongside the new, and each editor will choose his tool. Uziel302 (talk) 15:03, 16 October 2022 (UTC)[reply]
@Uziel302, some of the current pages have a lot of corrections that I can't fix immediately, possibly because other editors have dealt with the low hanging fruit. The tool will possibly mean that complex typos will be skipped by several editors, but the aim of the project is to fix typos in one click, so perhaps typos that can't be fixed in one click are a lower priority. I am not suggesting that you try to recreate the old pages in the new tool. Time will tell if the tool has further features that need to be tweaked. Excluding typos which have already been fixed is a huge plus. TSventon (talk) 10:22, 17 October 2022 (UTC)[reply]

I much prefer the way "type" works on the "old" system. I use "type" a lot, and when I do, a dialog pops up with the typo prefilled. Then I usually don't have to actually type anything; I just tap on the form and the browser recognizes the typo and suggests possible correct spellings which I can select, and one of them is almost always the desired one. This capability may depend on your browser; it works great in Safari on an iPad. This won't work on the "new" system, where the typing takes place in a form prefilled with a correct spelling. I can tap/Select/Replace to get possible replacements based on the correct spelling, but that's vastly inferior to generating options from the actual typo, and more inconvenient than a single tap . (I could copy the typo to the form and get the suggestions, but I'd be more likely to just type the correct word rather than go to the trouble.)  MANdARAXXAЯAbИAM  21:45, 16 October 2022 (UTC)[reply]

@Uziel302: thank you for making a new version! So far it works well and I have no issues but will let you know as I use it more. Also can you share the new tool to the English Wikivoyage editors at the pub as the typo correcting tool is very beneficial for that wiki as well. Gizza (talkvoy) 23:05, 16 October 2022 (UTC)[reply]
@DaGizza, I plan to add option in the tool to select a project, when we will have Wikivoyage support, I will share there. Uziel302 (talk) 23:41, 16 October 2022 (UTC)[reply]
@DaGizza, I added Wikivoyage to the tool, can you test, and if helpful, share with the community there? Uziel302 (talk) 16:58, 13 May 2023 (UTC)[reply]
Yes, it seems more helpful for "type" to show the original text rather than a proposed correction which the editor has just considered and rejected. Is it also worth including surrounding words, in case the correction involves them too? I might find, for example, that "for khandles" needs changing to "fork handles". This can't be done with either old or new "type", because "for" is (quite reasonably) not detected as a typo. Certes (talk) 23:55, 16 October 2022 (UTC)[reply]
@Certes, @Mandarax, I added type button with the same old behavior.
For the surrounding issues, I made the whole context paragraph editable. I'm a bit worried that it will cause some edge cases bugs in terms of encoding differences etc., so please follow carefully the actual result of editing context. Uziel302 (talk) 11:27, 17 October 2022 (UTC)[reply]
Thanks for creating this - it seems much more efficient. I used it last night and it worked fine, however this morning it is showing me the same typos as last night. Any idea what's causing that? Bellowhead678 (talk) 07:58, 18 October 2022 (UTC)[reply]
@Bellowhead678, this morning I changed status of served typos to new. If you clicked skip or dismiss, the status wouldn't stay as "served", served was only created to prevent serving the same typo to multiple users at the same time, and if no action was made, those will turn back to the "new" status and will be served in the tool. Thanks for using the tool and for all the hard work through the years. Uziel302 (talk) 11:27, 18 October 2022 (UTC)[reply]
I see another issue, some typos are failing to be found when trying to replace, but are found on the initial check. This is why they keep showing up, nothing changes their status. Working on a fix. Uziel302 (talk) 11:40, 18 October 2022 (UTC)[reply]
Uziel302, presumably that is why I am getting messages like the one below. It is easy to attempt a lot of these withoutnoticing if you don't look at the listing at the bottom of the page.
adoptations => adaptations in World Sailing
failed finding typo in article TSventon (talk) 12:11, 18 October 2022 (UTC)[reply]
@TSventon, @Bellowhead678, I fixed the bug that was causing "failed finding typo in article". Uziel302 (talk) 12:41, 18 October 2022 (UTC)[reply]
@Bellowhead678, there was another bug, template button didn't work properly, I added a fix and verified that now it acts as expected. I will go over the edits of the tool with template in the summary (which didn't have the bug) and fix. Uziel302 (talk) 16:02, 18 October 2022 (UTC)[reply]

Too many non-typos or wrong suggestions.

[edit]

I just went through a few (not all) of the recent edits of Siuhl10 (talk · contribs), who used the script to introduce numerous typos. See my recent contributions for some. "Toadish" is not a typo, neither should "substatute" (a sub-statute) be changed to "substitute". "theree parliaments" means "three parliaments", not "there parliaments". Typo fixing must be done with a less than 1 in 1000 error rate, otherwise it is disruptive and unhelpful. I realise that this is not the tool's fault, but user error; still, it should be more difficult to introduce so many mistakes using the tool. —Kusma (talk) 15:34, 13 January 2023 (UTC)[reply]

If "readd" and "reaward" are considered typos, the script should offer "re-add" and "re-award", not just "read" and "reward" as possible fixes. —Kusma (talk) 15:49, 13 January 2023 (UTC)[reply]
@Kusma, the responsibility lies with the user, not with the tool, to make sure the edits are okay. The tool doesn't decide which of the possible replacements is the most valid. — Qwerfjkltalk 15:59, 13 January 2023 (UTC)[reply]
I blame the tool for not giving sufficient context (with just a few words, how do you see you're in a quote or what language the text is in?). I have no idea why it would suggest to change "sacular" to "scapular" when it is probably "saccular". It would be better to leave rare words like these alone. —Kusma (talk) 16:03, 13 January 2023 (UTC)[reply]
@Kusma, the toolforge tool doesn't have this problem. — Qwerfjkltalk 16:42, 13 January 2023 (UTC)[reply]
I certainly wouldn't blame this one on the tool. (But it could omit anything that has "sic" close to it to avoid people misusing it). —Kusma (talk) 16:29, 13 January 2023 (UTC)[reply]
I think those are perfectly sensible suggestions, even if wrong in this instance. Sadly, no tool can be intelligent enough to get the right fix all the time, and this is the best we have. Certes (talk) 19:58, 13 January 2023 (UTC)[reply]
Turning an ambiguous typo into the wrong word is significantly worse than leaving the typo alone. I do not think it is sensible to offer a single-click suggestion in such cases, as it makes it too easy to do a lot of damage with the tool. And "toadish" is not a typo. —Kusma (talk) 20:49, 13 January 2023 (UTC)[reply]
Turning an ambiguous typo into the wrong word is significantly worse than leaving the typo alone, but the tool does not do that. The tool makes suggestions, a few of which would make the article worse, just as AWB and other tools do. On balance, it's a big net positive. Certes (talk) 22:08, 13 January 2023 (UTC)[reply]
Kusma, in the end, Wikipedia is easily edited so editors easily write mistakes. The tool doesn't make edits, editors do. The tool indeed makes it easier, like the whole Wikipedia made it easier to edit online encyclopedia, and the price is that sometimes some editors make mistakes. I agree that we should have more context, and we did some progress in that direction by creating the toolforge tool, the recent list wasn't through that tool due to lack of time and some technical issues on my side. Uziel302 (talk) 14:02, 19 January 2023 (UTC)[reply]
Of course it is individual editors' responsibility to not introduce mistakes, I just think it should not be too easy to make many mistakes quickly (that's why AWB access is limited). The tool is super easy to use, which is quite nice (I just dismissed a lot of false positives, most of them in foreign language text, and fixed a handful of typos with it), I just think it could be even better, and I'm wondering how to discourage people from misusing it. —Kusma (talk) 19:21, 19 January 2023 (UTC)[reply]
Kusma, thanks for your feedback and contribution, I now uploaded a new batch of typos to the toolforge tool, which has significantly more context. Looking forward to hearing your feedback. As of limiting the access, it was offered before, it is on my tasks list, hope to get to it soon. Thanks. Uziel302 (talk) 18:26, 28 January 2023 (UTC)[reply]
Thanks for the upload. I've worked through some of the new batch and made several useful fixes, most of which were correctly suggested. However there are still lots of false positives and most of them are among foreign-language text.I often see something like "... nybat bul juhuryb aramir aramis ausibør pisitʌnʌgel usibɛl...". No thanks: I've no reason to believe that changing aramir to aramis would suddenly make that comprehensible to our readers. Can we exclude "typos" where most of the surrounding "words" aren't in the dictionary either? Certes (talk) 00:14, 29 January 2023 (UTC)[reply]
Great idea. Obviously it would require some coding on my end, I hope to get to it soon. Uziel302 (talk) 16:50, 30 January 2023 (UTC)[reply]

Certes, I recently implemented the following logic: if a line has more than one word not appearing in Wikipedia in three articles, that line won't add suspects to the list. I uploaded new lists, also added projects and languages (some of the list is still in progress). In Wikivoyage it was very helpful to filter as I said, because they have many names of places and phrases in foreign languages. Would love to hear any feedback. Uziel302 (talk) 16:57, 13 May 2023 (UTC)[reply]

That looks better! I'm busy this week but I just looked at ten suggestions and five of them were actual typos I could fix, which is an improvement. It seems to have done a good job of filtering out the large chunks of non-English prose which used to come up. Removing suggestions with just one other non-word in the line seems a bit severe, but it's a great start and could easily be relaxed later. Certes (talk) 08:17, 14 May 2023 (UTC)[reply]

The script is mostly wrong

[edit]

Why is the script suggesting such weird suggestions? Snagstaxi (talk) 02:05, 3 April 2023 (UTC)[reply]

Making such weird suggestions as what? For example? What do you mean? How so? In what way? Etc. — JohnFromPinckney (talk / edits) 08:50, 3 April 2023 (UTC)[reply]
Sometimes, multiple words are equally close to the typo. Currently, the script suggests the first such word alphabetically. For example, if an article contains "kiled", the suggested fix is "kidel" rather than "killed". As discussed above, we are considering changing that algorithm, to suggest common words in preference to an uncommon ones. Certes (talk) 14:53, 3 April 2023 (UTC)[reply]

Lists in foregn language

[edit]

List of compositions by Alexander Agricola is a list of songs mostly in Latin or fourteenth century French. Is there a good way of excluding it from the tool. I have just dismissed the suggestion actendant → attendant, which would be the modern French spelling, but not the fourteenth century French spelling. TSventon (talk) 12:32, 14 May 2023 (UTC)[reply]